An embedded and charset-unspecified text was scrubbed... Name: not available URL: <https://stat.ethz.ch/pipermail/r-sig-mixed-models/attachments/20090113/2851606d/attachment.pl>
Zero variance and Std. Dev. using lmer?
4 messages · Luciano La Sala, Renwick, A. R., ONKELINX, Thierry +1 more
I asked a similar question and got a good reponse last year. Follow the link below: https://stat.ethz.ch/pipermail/r-sig-mixed-models/2008q3/001245.html -----Original Message----- From: r-sig-mixed-models-bounces at r-project.org [mailto:r-sig-mixed-models-bounces at r-project.org] On Behalf Of Luciano La Sala Sent: 13 January 2009 14:16 To: r-sig-mixed-models at r-project.org Subject: [R-sig-ME] Zero variance and Std. Dev. using lmer? Dear R-people: I have run a GLMM (using lmer) with the fixed and random effects detailed below. Oddly I think, I get zero variance and Std. Dev. values. How is that possible? Does it mean that the RE "NestID" is not helping to account for autocorrelation among sibling chicks at the nest level? Or is this a small sample size problem? As well, I ran an ordinary logistic regression using he exact same fixed variables, and I got the exact same AIC and BIC values and estimates for fixed effects, error, z value, and Pr(>|z|). Does this support the idea that the GLMM with RE for NestID is not necessary at all? Look forward to hearing from you. Cheers for now. Luciano GENERALIZED LINEAR MIXED MODEL WITH RANDOM INTERCEPT model <- lmer(Death10~HO+ClutchSize+SibComp+Yr+(1|NestID),family=binomial,1) Generalized linear mixed model fit by the Laplace approximation Formula: Death10 ~ HO + ClutchSize + Sibcomp + yr + (1 | NestID) Data: 1 AIC BIC logLik deviance 242.2 268.5 -113.1 226.2 Random effects: Groups Name Variance Std. Dev. NestID (Intercept) 0 0 Number of obs: 198, groups: NestID, 104 Fixed effects: Estimate Std. Error z value Pr(>|z|) (Intercept) -1.2239 0.5114 -2.3934 0.0167 * HOSecond -0.6910 0.8928 -0.7739 0.4390 HOThird 0.6768 1.0327 0.6554 0.5122 ClutchSizeTwo-eggs 1.3961 0.5864 2.3809 0.0173 * ClutchSizeThree-eggs 0.3958 0.5843 0.6773 0.4982 SibcompAbsent 1.7804 0.9140 1.9479 0.0514 . yr2007 -0.8299 0.3423 -2.4245 0.0153 * --- Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 Correlation of Fixed Effects: (Intr) HOScnd HOThrd CltchSzTw- CltchSzTh- SbcmpA HOSecond -0.034 HOThird -0.031 0.837 CltchSzTw-g -0.830 -0.069 -0.022 CltchSzThr- -0.816 -0.088 -0.107 0.785 SibcmpAbsnt 0.052 -0.904 -0.836 -0.018 -0.050 yr2007 -0.233 0.145 0.133 0.025 -0.050 - 0.224 _______________________________________________ R-sig-mixed-models at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models The University of Aberdeen is a charity registered in Scotland, No SC013683.
Dear Luciano,
Your variables are strongly correlated. Sibling competition is only
absent when the clutch size is one. Likewise the hatching order can only
be three if the clutch size is three. This could cause numberical
instability of your model. So I suggest that you simplify your model.
What results do you get with this model: lmer(Death10 ~ ClutchSize + Yr
+ (1|NestID), family = binomial)
HTH,
Thierry
------------------------------------------------------------------------
----
ir. Thierry Onkelinx
Instituut voor natuur- en bosonderzoek / Research Institute for Nature
and Forest
Cel biometrie, methodologie en kwaliteitszorg / Section biometrics,
methodology and quality assurance
Gaverstraat 4
9500 Geraardsbergen
Belgium
tel. + 32 54/436 185
Thierry.Onkelinx at inbo.be
www.inbo.be
To call in the statistician after the experiment is done may be no more
than asking him to perform a post-mortem examination: he may be able to
say what the experiment died of.
~ Sir Ronald Aylmer Fisher
The plural of anecdote is not data.
~ Roger Brinner
The combination of some data and an aching desire for an answer does not
ensure that a reasonable answer can be extracted from a given body of
data.
~ John Tukey
-----Oorspronkelijk bericht-----
Van: r-sig-mixed-models-bounces at r-project.org
[mailto:r-sig-mixed-models-bounces at r-project.org] Namens Luciano La Sala
Verzonden: dinsdag 13 januari 2009 15:16
Aan: r-sig-mixed-models at r-project.org
Onderwerp: [R-sig-ME] Zero variance and Std. Dev. using lmer?
Dear R-people:
I have run a GLMM (using lmer) with the fixed and random effects
detailed
below. Oddly I think, I get zero variance and Std. Dev. values.
How is that possible? Does it mean that the RE "NestID" is not helping
to
account for autocorrelation among sibling chicks at the nest level?
Or is this a small sample size problem?
As well, I ran an ordinary logistic regression using he exact same fixed
variables, and I got the exact same AIC and BIC values and estimates for
fixed effects, error, z value, and Pr(>|z|).
Does this support the idea that the GLMM with RE for NestID is not
necessary
at all?
Look forward to hearing from you.
Cheers for now.
Luciano
GENERALIZED LINEAR MIXED MODEL WITH RANDOM INTERCEPT
model <-
lmer(Death10~HO+ClutchSize+SibComp+Yr+(1|NestID),family=binomial,1)
Generalized linear mixed model fit by the Laplace approximation
Formula: Death10 ~ HO + ClutchSize + Sibcomp + yr + (1 | NestID)
Data: 1
AIC BIC logLik deviance
242.2 268.5 -113.1 226.2
Random effects:
Groups Name Variance Std. Dev.
NestID (Intercept) 0 0
Number of obs: 198, groups: NestID, 104
Fixed effects:
Estimate Std.
Error
z value Pr(>|z|)
(Intercept) -1.2239 0.5114
-2.3934 0.0167 *
HOSecond -0.6910 0.8928
-0.7739 0.4390
HOThird 0.6768 1.0327
0.6554 0.5122
ClutchSizeTwo-eggs 1.3961 0.5864
2.3809
0.0173 *
ClutchSizeThree-eggs 0.3958 0.5843
0.6773
0.4982
SibcompAbsent 1.7804 0.9140
1.9479 0.0514 .
yr2007 -0.8299 0.3423
-2.4245 0.0153 *
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Correlation of Fixed Effects:
(Intr) HOScnd HOThrd CltchSzTw-
CltchSzTh- SbcmpA
HOSecond -0.034
HOThird -0.031 0.837
CltchSzTw-g -0.830 -0.069 -0.022
CltchSzThr- -0.816 -0.088 -0.107 0.785
SibcmpAbsnt 0.052 -0.904 -0.836 -0.018
-0.050
yr2007 -0.233 0.145 0.133 0.025
-0.050 - 0.224
_______________________________________________
R-sig-mixed-models at r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
Dit bericht en eventuele bijlagen geven enkel de visie van de schrijver weer
en binden het INBO onder geen enkel beding, zolang dit bericht niet bevestigd is
door een geldig ondertekend document. The views expressed in this message
and any annex are purely those of the writer and may not be regarded as stating
an official position of INBO, as long as the message is not confirmed by a duly
signed document.
On Tue, 13 Jan 2009, Luciano La Sala wrote:
Dear R-people: I have run a GLMM (using lmer) with the fixed and random effects detailed below. Oddly I think, I get zero variance and Std. Dev. values. How is that possible? Does it mean that the RE "NestID" is not helping to account for autocorrelation among sibling chicks at the nest level? Or is this a small sample size problem? model <- lmer(Death10~HO+ClutchSize+SibComp+Yr+(1|NestID),family=binomial,1)
Either, I would think. Have you performed a simple test for extrabinomial variation? eg Tarone test or Chi-square [latter is just contingency Chi-square test for NestID x Death10] David Duffy.
| David Duffy (MBBS PhD) ,-_|\ | email: davidD at qimr.edu.au ph: INT+61+7+3362-0217 fax: -0101 / * | Epidemiology Unit, Queensland Institute of Medical Research \_,-._/ | 300 Herston Rd, Brisbane, Queensland 4029, Australia GPG 4D0B994A v